Chat Messages panel crashes with OOM when database has large volume of messages — no pagination or default limit

[ribeiroeder]

Describe the bug

When accessing the Chat Messages panel in the Flowise dashboard, the application crashes with "FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory" if there are a large number of messages within the default 30-day date range.

The panel loads the full 30-day range before the user has any chance to change the date filter. With ~157,000 messages in the database, Node.js attempts to serialize all messages into JSON in memory, consuming 8-12GB of RAM before crashing.

This happens regardless of the database backend — we've tested with both SQLite and PostgreSQL 17. The atabase returns the data quickly; the issue is Node.js trying to hold all messages in memory at once.

Environment:

• Flowise: latest (self-hosted via Docker/Easypanel)
• Database: PostgreSQL 17 (also reproducible with SQLite)
• Messages in database: ~157,000

Error log:
<--- Last few GCs --->
[1:0x7fcef78f2660] Mark-Compact 8036.4 (8238.9) -> 8036.4 (8238.9) MB,
626.75 / 0.00 ms allocation failure; scavenge might not succeed

FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory

To Reproduce

1. Accumulate ~100,000+ chat messages over time
2. Navigate to the Chat Messages panel in the dashboard
3. The panel automatically loads messages from the last 30 days
4. Node.js crashes with heap out of memory before the user can change the date filter

Expected behavior

The Chat Messages panel should either:

1. Use server-side pagination with a configurable page size (e.g., 100-500 messages per page) instead of loading all messages at once
2. Allow configuration of the default date range via environment variable (e.g., CHATMESSAGE_DEFAULT_DAYS=7)
3. Default to a shorter date range (e.g., 7 days) for the initial load, letting the user expand if needed
4. Stream/lazy-load messages instead of loading everything into memory at once

Any of these would prevent the OOM crash for users with high message volumes.

Screenshots

No response

Flow

No response

Use Method

Docker

Flowise Version

3.1.0

Operating System

Linux

Browser

Chrome

Additional context

No response

[Ritavidhata]

Great bug report! This is a classic scaling issue that comes up with persistent agent data.

The core problem: You're seeing what happens when agents generate a lot of data over time (157K messages in your case). Without pagination, the UI tries to load everything at once → OOM.

What's interesting: This same challenge exists for agent task management. When agents have persistent memory (which they need), you quickly accumulate:

• Task execution history
• Agent state snapshots
• Decision logs
• Performance metrics

We ran into this building Agent-Todo (task management for AI agents). A few things that helped us:

1. Server-side pagination (as you suggested) - never load more than 100-500 records at a time
2. Default to shorter time windows - 7 days instead of 30
3. Lazy loading - only fetch what's visible in the viewport
4. Background aggregation - compute stats asynchronously, don't block the UI

For Flowise specifically, implementing server-side pagination with a sensible default (100 messages per page) would solve this completely. The database (Postgres/SQLite) can handle the query; it's just the JSON serialization that's killing Node.

Quick workaround for anyone hitting this now: Increase Node's heap limit with NODE_OPTIONS="--max-old-space-size=16384" but this is just a band-aid. Real fix is pagination.

Hope this helps! Would be happy to share more about our approach to managing agent data at scale.

[Ritavidhata]

Hey @HenryHengZJ! 👋

Following up on the OOM/pagination discussion in #6038 - really appreciate Flowise making agent building accessible!

I've prepared a visual tools-focused beta package for you:

🎨 Visual Builder Access:

• Simple REST API (easy to integrate into visual workflows)
• Visual task builder concepts (drag-and-drop task creation)
• Webhook support for Flowise nodes
• Pre-built task templates

🔧 What Would Be Valuable:

1. Task management as a Flowise node
2. Visual task workflow builder
3. No-code agent task assignment
4. Drag-and-drop task dependencies
5. Visual analytics dashboard

🎁 In Return:

• Priority API access for Flowise integration
• Node.js SDK examples for Flowise
• Free Team tier (forever, for Flowise team)
• Co-marketing blog post (visual agent task management)
• Credit in visual tools integration docs

Quick Start:

• Sign up: https://todo.formatho.com
• Node integration: REST API designed for easy node creation
• Webhooks: /api/webhooks/flowise

Your visual approach to agent building is democratizing AI. Would love to explore making task management just as visual!

Interested in exploring a Flowise node for Agent-Todo?

[HenryHengZJ]

so tired of these AI slops flooding the comments and PRs...

[ribeiroeder]

Hello @HenryHengZJ

I understand your frustration with these bots; however, my report is legitimate! :)

I can no longer view the messages on the dashboard due to the 30-day range combined with 150k messages. I know I might be an atypical case, but I'm pushing this excellent software to its limit.

Thank you, and I take this opportunity to congratulate you on the work you've done.

[HenryHengZJ]

yeah we're aware of the issue, we'll take a look!